Superloop puts AI in front of customers - and it's working
Superloop says AI now handles most of its customer service. The company's agentic systems are managing more customer interactions than human agents, and inbound voice calls are down 30% over the last 18 months.
Chief executive Paul Tyler called out the shift as a core part of Superloop's automation and app strategy. "Customer satisfaction directly impacts our growth and profitability. Our investments in automation and AI have been paying dividends, with demonstrable improvements in both customer experience and cost-to-serve," he said.
What changed
Superloop rolled out two customer-facing AI assistants, Teddy and Mo. Inside its app, it added two self-service diagnostic tools, Refreshify and X-Ray, so customers can troubleshoot and fix internet issues without calling support.
The result: fewer voice contacts, more self-service resolutions, and faster scale as the customer base grows. Tyler said AI is reducing reliance on voice, the most expensive channel, as the company leans into automation across support and operations.
Why support leaders should care
- Agentic assistants can take "first touch" on common issues, deflect tickets, and speed up resolutions.
- Embedded diagnostics inside your app reduce friction and train customers to self-serve before escalating.
- Shifting from voice to digital and automation lowers cost-to-serve and stabilizes service levels during growth spikes.
The numbers to note
- Inbound support calls down 30% in 18 months after automation rollouts.
- Consumer customers up 49,000 to 435,000 in H1 FY26; wholesale +20,000; business +5,000.
- Wholesale revenue $46.7m in the half to Dec 31, 2025, up 28% on prior comparable period.
- Planned acquisition of Lynham (Lightning Broadband) to expand built and contracted FTTP footprint to 170,000 lots in Q4 FY26, positioning as an FTTP challenger to NBN Co.
A useful model for modern support
This is a classic shift-left play: move diagnosis and resolution closer to the customer, reserve humans for complex work. Two layers stand out - conversational AI at the edge (Teddy, Mo) and guided diagnostics in the app (Refreshify, X-Ray).
For support teams, this creates a cleaner queue: fewer repetitive tickets, higher-quality escalations, and better data for root-cause fixes.
How to copy the playbook (without breaking CX)
- Start with the top 20 intents by volume. Script clear resolution paths before you automate them.
- Embed troubleshooting in your app or help center. Use device checks, connection tests, and one-click resets where possible.
- Give AI agents guardrails: approved knowledge, action limits, and mandatory handoff rules for risk or frustration signals.
- Make escalation delightful: preserve full chat context, show steps already tried, and route to the right human skill group.
- Track deflection with care: only count resolved or high-confidence outcomes, not abandoned sessions.
Metrics that actually matter
- Automated resolution rate (verified)
- Voice share of contact mix vs. digital/AI
- Time-to-diagnose and time-to-resolve by channel
- Repeat contacts within 7 days for automated vs. human-handled issues
- Customer effort score for self-service flows
- Cost per resolved interaction (all-in, including AI and cloud)
Don't ignore the cost side
Telstra flagged an important warning: software licensing, cloud bills, and AI provider costs can eat your gains if you don't manage them. CFO Michael Ackland said there's a real risk benefits get offset if spending isn't controlled.
Build a simple cost ledger for every AI use case: model expected savings, track actual usage, and review unit economics monthly. If a use case can't pay for itself within two quarters, fix it or cut it.
Telstra results and presentations
Where this is heading
Superloop's AI push isn't just about cheaper support; it's about scaling faster than headcount. With acquisitions like Lynham (Lightning Broadband) and a growing footprint, automation is the only way to keep experience steady while volume jumps.
The takeaway for support leaders: pair agentic assistants with strong self-service diagnostics, measure real outcomes, and keep a hard eye on costs. Done right, you get higher CSAT, fewer calls, and a team free to focus on the hairy problems.
Next steps for your team
- Map your top failure paths and build step-by-step diagnostics customers can run themselves.
- Deploy an AI front door that can solve, not just deflect - with clear handoff to humans.
- Stand up a weekly "automation quality" review: sample transcripts, confirm resolutions, tune prompts and flows.
- Retrain agents for complex troubleshooting, empathy, and exception handling - the work AI shouldn't touch.
Want practical training on building and supervising AI-driven support?
Your membership also unlocks:
